Resoneo says ChatGPT responses began referencing about 20% fewer websites after what it identifies as the early-March transition to GPT-5.3 Instant.
The analysis comes from the French SEO consultancy and draws on data from Meteoria, an AI visibility-tracking platform that monitored 400 prompts daily over 14 weeks, producing 27,000 comparable responses.
Average unique domains per response dropped from 19 before the transition to 15 after. Average unique URLs per response fell from 24 to 19.
The URLs-per-domain ratio remained at 1 throughout the tracking period. The data suggests ChatGPT isn’t visiting as many sites per response, but it’s going just as deep into each one.
Fewer domains now share the same citation surface in each response, meaning the sites that do get cited take up a larger share of each answer.
Server Logs Back Up The Pattern
Independent log analysis from Jérôme Salomon at Oncrawl supports the findings. Tracking ChatGPT-User bot activity across multiple websites, his data shows crawl volume has settled at a lower level. Some pages aren’t being crawled at all anymore, and the crawl frequency for pages still being visited has dropped.
Resoneo links the change to ChatGPT’s default experience now being driven more heavily by GPT-5.3 Instant, which the company says triggers fewer web searches and citations than earlier behavior. Oncrawl’s server log data shows the lower crawl pattern over the same period.
Earlier data has shown how AI platforms cite sources differently from traditional search. An SE Ranking analysis of 129,000 domains found that referring domains were the strongest predictor of the likelihood of ChatGPT citation, with a threshold effect at 32,000 referring domains.
A Search Atlas report similarly showed low overlap between Google rankings and ChatGPT citations, with median domain overlap around 10-15%.
Why This Matters
A 20% drop in cited domains per response means fewer websites competing for visibility inside each ChatGPT answer. The total citation surface shrank, but the sites that kept appearing maintained their same crawl depth.
For anyone tracking referral traffic from ChatGPT, the early-March model transition is a date range worth checking in your analytics.
Looking Ahead
Resoneo’s analysis notes that GPT-5.4 Thinking reintroduces search fan-outs and uses site: operators to target trusted domains, but these behaviors weren’t captured in the quantitative dataset, which covers GPT-5.3 Instant and below.
Whether the citation surface continues to narrow or widens again with newer models isn’t yet clear.
